News for hackers. Stuff that matters.

A comment on GPT-4.5. Parent comment

Re:GPT-4.5 by turnsout on Thursday February 27, 2025 @05:05PM (#43199045)
If you like absurdist humor, go into the OpenAI playground, select 3.5-Turbo, and dial up the temperature to the point where the output devolves into garbled text after 500 tokens or so. The first ~200 tokens are in the freaking sweet spot of humor.
[ Parent | on HN ]
Re:GPT-4.5 by rl3 on Thursday February 27, 2025 @05:31PM (#43199275)
Maybe it's rose-colored glasses, but 3.5 was really the golden era for LLM comedy. More modern LLMs can't touch it.

Just ask it to write you a film screenplay involving some hard-ass 80s/90s action star and someone totally unrelated and opposite of that. The ensuring unhinged magic is unparalleled.

[ Parent | on HN ]
Re:GPT-4.5 by jcims on Thursday February 27, 2025 @05:37PM (#43199325)
I built a little AI assistant to read my calendar and send me a summary of my day every morning. I told it to roast me and be funny with it.

3.5 was *way* better than anything else at that.

[ Parent | on HN ]
Re:GPT-4.5 by rl3 on Friday February 28, 2025 @01:07AM (#43202242)
Yeah, I think the fact its "mind" so to speak was more fragmented and unpredictable was almost a boon for that purpose.
[ Parent | on HN ]
Re:GPT-4.5 by sebastiennight on Friday February 28, 2025 @02:48AM (#43202876)
Ah, I'd love to have that kind of daily recap... mind sharing some of the code (or even just the prompt?)
[ Parent | on HN ]
Re:GPT-4.5 by rl3 on Friday February 28, 2025 @01:06AM (#43202233)
>The ensuring unhinged magic is unparalleled.

Oops: ensuing*

[ Parent | on HN ]
Re:GPT-4.5 by amarcheschi on Thursday February 27, 2025 @05:26PM (#43199234)
Could someone post an example?
[ Parent | on HN ]